Introduction

NEFA have been shown to both stimulate and impair insulin secretion depending on the duration of beta cell exposure [1]. Prolonged (>24 h) exposure is inhibitory and involves oxidative stress, endoplasmic reticulum stress and inflammation [1].

Lipotoxic effects on the beta cell, i.e. the decrease in beta cell function and mass induced by chronically elevated NEFA, play a role in the pathogenesis of type 2 diabetes, at least in predisposed individuals [1]. Previously, we demonstrated that oxidative stress mediates fat-induced beta cell dysfunction in vivo in rats [2] and in humans [3]. Oxidative stress activates inhibitor of (nuclear factor) κB kinase β (IKKβ), which, by phosphorylating the inhibitor of (nuclear factor) κB α (IκBα), activates the transcription factor NFκB. The effect of IKKβ/NFκB on beta cell function is controversial. NFκB is important for cell survival and there are reports that NFκB is beneficial for glucose-stimulated insulin secretion [4], unless activated by cytokines [5]. However, IKKβ which induces serine phosphorylation of IRSs [6] in addition to activating NFκB, decreases beta cell function [7]. It is also controversial whether fat activates IKKβ/NFκB in beta cells: although fatty acids did not activate NFκB in INS-1 or primary rat beta cells in one study [8], lipotoxicity was associated with NFκB activation, and palmitate-induced apoptosis was inhibited by an IKKβ inhibitor in INS-1 beta cells in another study [9]. The in vivo effect of fat on beta cell IKKβ/NFκB has not been investigated previously.

To address the role of IKKβ in fat-induced beta cell dysfunction in vivo, rats were infused i.v. with fat to elevate plasma NEFA 50–100% (elevation seen in obesity [10]) with or without the IKKβ inhibitor [11] salicylate for 48 h. Monounsaturated fat (oleate or olive oil) was infused, as in our previous study showing a role of oxidative stress in beta cell dysfunction [2]. Although oleate has been found to protect beta cells against palmitate-induced toxicity in in vitro studies [12], the prolonged effect of oleate by itself on beta cell function is mainly inhibitory [2, 13]. After 48 h infusion, beta cell function was evaluated in vivo using hyperglycaemic clamps, or ex vivo in isolated islets. We also used the specific IKKβ inhibitor 4(2′-aminoethyl)amino-1,8-dimethylimidazo(1,2-a)quinoxaline; BMS-345541 (BMS) [14] in hyperglycaemic clamps performed after 48 h oleate infusion in mice. In addition, we exposed rat islets for 48 h to oleate with or without salicylate or BMS in vitro. Last, we performed oleate infusion in beta cell-specific IKKβ-deficient (Ikkb Δbeta cell) mice, followed by evaluation of beta cell function in vivo using hyperglycaemic clamps or ex vivo in isolated islets. In all these models, IKKβ inhibition prevented fat-induced beta cell dysfunction.

Methods

Animals

All procedures were approved by the Animal Care Committee of the University of Toronto and conducted according to the Canadian Council on Animal Care Guidelines and the appropriate sample sizes were determined prior to the studies (see also Electronic supplementary material [ESM]). Female Wistar rats (250–300 g, corresponding to 9–11 weeks of age, Charles River, Senneville, QC, Canada) were used, as in our previous studies [2, 15]. Female C57BL/6 mice (22–25 g, corresponding to 14–16 weeks of age, Jackson Laboratory, Bar Harbor, ME, USA) and male RIP2 (rat insulin 2 promoter)-Cre recombinase (Cre) positive or negative Ikkb floxed mice were also used [16]. Female wild-type C57BL/6 mice underwent the BMS infusion studies. For studies in beta cell-specific Ikkb-null mice, Ikkb F/F mice on a C57BL/6 background were crossed with RIP2-Cre mice (also on a C57BL/6 background) obtained from Jackson Laboratory to generate RIP2-Cre positive or RIP2-negative Ikkb F/+ mice, which were interbred to generate RIP2-Cre−:Ikkb F/F (floxed controls) and RIP2-Cre+:Ikkb F/F (Ikkb Δbeta cell) mice. These mice were used for experiments at 11–13 weeks of age. Cre-mediated recombination was confirmed by PCR [17].

Surgery and i.v. infusion

Surgery was performed under general anaesthesia (isoflurane, to effect). Carotid artery and jugular vein cannulation of rats has been described previously [2, 15]. After 3 days’ post-surgical recovery, the rats were randomised to the following 48 h infusions: (1) NaCl (154 mmol/l; 5 μl/min) as control; (2) fat, either oleate (1.3 μmol/min) prepared as in our previous studies [2, 15] or 20% (vol./vol.) olive oil (OLO; 5.5 μl/min) prepared as in Dobbins et al [18]; (3) oleate or OLO + salicylate (0.7 μmol kg−1 min−1, the dose that reversed insulin resistance in Park et al [19]); or (4) salicylate only. Olive oil is a triacylglycerol mixture containing 75% oleate, and 16% saturated fat. Heparin was added to olive oil to a final concentration of 50 U/ml to activate lipoprotein lipase, which releases NEFA and glycerol from the triacylglycerol mixture of olive oil. We have shown that BSA, the vehicle for oleate, has no effect on insulin secretion [15, 20]; heparin [21] and glycerol [22] also have no effect. A two-step hyperglycaemic clamp or islet isolation was performed after the infusion period.

Mouse jugular vein cannulation is described in Koulajian et al [20]. In mice, oleate (0.4 μmol/min)/equivolume saline was infused for 48 h [20], starting 4–5 days after surgery with or without BMS (0.12 mmol kg−1 day−1) [14]. After 48 h infusion, mice received hyperglycaemic clamps. Sampling was through the tail vein.

Hyperglycaemic clamp

Insulin secretion was determined by measuring plasma insulin and C-peptide during a two-step (13 and 22 mmol/l) hyperglycaemic clamp in rats [2] and a one-step (22 mmol/l) hyperglycaemic clamp in mice [20]. Insulin sensitivity (M/I, where M is glucose metabolism and I is plasma insulin) and beta cell function (disposition index [DI]; units are μmol kg−1 min−1 glucose/pmol insulin × nmol C-peptide) were assessed as described in the ESM.

Hyperinsulinaemic–euglycaemic clamp

Insulin sensitivity was determined using the gold-standard technique, i.e. hyperinsulinaemic–euglycaemic clamp [19].

Ex vivo studies

Islets of the in vivo infused rats and Ikkb Δbeta cell mice were isolated and insulin secretion studies were performed as described previously [2]. ELISA for phosphorylated IκBα and active nuclear NFκB (Active Motif, Carlsbad, CA, USA), reactive oxygen species (ROS) measurements (see ESM) and RT-PCR were also performed in rat islets. Total rat islet RNA was prepared as described previously [23, 24] according to the manufacturer’s instructions (Qiagen, Hombrechtikon, Switzerland), and was reverse transcribed using random hexamers. Mouse primers for Il1b, Tnfa (also known as Tnf), monocyte chemoattractant protein 1 (Mcp1; also known as Ccl2), IL-1 receptor antagonist (Il1ra; also known as Il1rn), Tgfb (also known as Tgfb1), the macrophage marker Cd68, and cyclooxygenase (Cox)2 (also known as Ptgs2) (see ESM Table 1) from Applied Biosystems (Foster City, CA, USA) were used and quantitative PCR was done with a fluorescein amidite (FAM)-based reference dye using commercial TaqMan gene expression assays and the 7500 Fast Real-Time PCR System according to the manufacturer’s protocol (Applied Biosystems). Changes in mRNA expression were calculated using difference in Ct values, and normalised to the housekeeping gene 18S [23, 24].

Studies in cultured islets

Rat islets were cultured for 48 h in RPMI 1640 without antioxidants, containing 0.4 mmol/l oleate in 0.5% NEFA-free BSA with or without 0.25 mmol/l salicylate [7] or 3 μmol/l BMS, a dose based on pilot dose–response studies. Islets were also cultured in control/oleate media with or without the COX-2 inhibitor SC-236 (10 μmol/l, based on Castaño et al [25]) or the COX-1 inhibitor SC-560 (100 μmol/l, based on Smith et al [26]). Thereafter, glucose-stimulated insulin secretion was assessed. Prostaglandin E2 (PGE2) in the medium was measured via ELISA (Enzo Life Sciences, Farmingdale, NY, USA).

Western blots

Western blots were performed as previously described [19] using antibody against phospho-Ser307-IRS-1 (Millipore, Billerica, MA, USA, 07-247, raised in rabbit, RRID:AB_310463, 1:500), total IKKβ (Cell Signaling Technology, Danvers, MA, USA, 2370, raised in rabbit, RRID:AB_2122154, 1:100), phospho-Thr172-5′ adenosine monophosphate-activated protein kinase (AMPK)α (Cell Signaling Technology, 2535, raised in rabbit, RRID:AB_331250, 1:1000), total AMPKα (Cell Signaling Technology, 2793, raised in mouse, RRID:AB_915794, 1:1000), total IκBα (Santa Cruz Biotechnology, Dallas, TX, USA, sc-371, raised in rabbit, RRID:AB_2235952, 1:250), α-actinin (Cell Signaling Technology, 3134S, raised in rabbit, RRID:AB_2223798, 1:1000), β-actin (Abcam, Cambridge, UK, ab6276, RRID:AB_2223210, raised in mouse, 1:10,000) or γ-tubulin (Sigma-Aldrich, St Louis, MO, USA, T6557, RRID:AB_477584, raised in mouse, 1:1000) (see also ESM). The antibodies had been validated in previous studies carried out in the laboratory (see ESM).

Plasma assays

Plasma glucose in rats was measured on a Beckman Analyser II (Beckman, Fullerton, CA, USA) and in mice was measured on a HemoCue Glucose 201 Analyser (HemoCue, Brea, CA, USA). Plasma NEFA were measured with an enzymatic colorimetric kit (Wako Industries, Neuss, Germany). Radioimmunoassays specific for rat/mouse insulin and C-peptide (Linco, St Charles, MO, USA) were used to determine their plasma concentrations. Plasma triacylglycerol levels were measured by a colorimetric kit (Boehringer Mannheim, Laval, QC, Canada).

Statistics

Data are means ± SEM. One-way non-parametric ANOVA for repeated measurements followed by Tukey’s test was used to compare treatments. Calculations were performed using SAS version 8.0 (Cary, NC, USA).

Results

Hyperglycaemic clamp in rats

Rats were infused i.v. with saline, oleate or OLO, with or without the IKKβ inhibitor salicylate. After 48 h infusion, levels of plasma NEFA were ~1.5-fold higher with oleate or OLO and triacylglycerol levels were elevated by OLO (Table 1). Oleate, OLO or salicylate did not affect plasma glucose or insulin (Table 1). Following the 48 h infusions, we evaluated beta cell function in vivo using two-step hyperglycaemic clamps (13 mmol/l and 22 mmol/l, Fig. 1a, b). The glucose infusion rate (GINF) necessary to maintain the clamp was lower with oleate or OLO, suggesting reduced insulin secretion and/or sensitivity (Fig. 1c, d). With oleate + salicylate or OLO + salicylate, GINF was similar to control (Fig. 1c, d).

Table 1 Plasma NEFA, triacylglycerol, glucose and insulin levels after 48 h infusions
Full size table
Fig. 1
figure 1

Plasma glucose (a, b), GINF (c, d), insulin (e, f) and C-peptide (g, h) during two-step hyperglycaemic clamps following 48 h of oleate or OLO infusion. (a, c, e, g) Rats were treated with: saline (n = 12); oleate at 1.3 μmol/min (n = 10); oleate + salicylate at 0.7 μmol kg−1 min−1 (n = 8); or salicylate only (n = 9). (b, d, f, h) Rats were treated with: saline (n = 12); OLO (20% olive oil infusate containing 50 U/ml heparin) at 5.5 μl/min (n = 7); OLO + salicylate (n = 11); or salicylate only (n = 9). Data are mean ± SEM. **p < 0.01 and ***p < 0.001 vs all, throughout the clamp. White circles, saline; black circles, oleate; black triangles, oleate + salicylate; white triangles, salicylate; black squares, OLO; black diamonds, OLO + salicylate

Full size image

Clamp plasma insulin was lower with oleate infusion and was restored with oleate + salicylate, whereas insulin was not different from control with OLO and OLO + salicylate (Fig. 1e, f). Clamp C-peptide showed the same pattern as insulin (Fig. 1g, h), indicating unchanged absolute insulin secretion in OLO-treated groups.

The insulin sensitivity index (M/I = GINF/insulin [27]) was not affected by oleate (Fig. 2a), as previously found in the same model [2, 15], and was also unaffected by salicylate. With OLO, M/I was reduced, indicating insulin resistance (Fig. 2b). The decrease in M/I was prevented by salicylate. The evaluation of beta cell function in vivo should take into account the ambient insulin sensitivity, as normal beta cells increase insulin secretion in response to insulin resistance along a hyperbola, characterised by a constant DI [28]. The DI is an established index of beta cell function [28] that we have previously validated in rodents [20, 24]. The DI was impaired in both oleate and OLO groups (Fig. 2c, d), indicating reduced beta cell function; the impairment was completely prevented by salicylate. Salicylate alone had no effect on DI.

Fig. 2
figure 2

M/I and DI during two-step hyperglycaemic clamps with and without infusion for 48 h of oleate (a, c) or OLO (b, d). See ESM Methods for calculation of these indices. The groups are as described for Fig. 1. Data are mean ± SEM. **p < 0.01 vs all groups at the same glucose concentration. White bars, saline; light grey bars, salicylate only. In (a, c): black bars, oleate; dark grey bars, oleate + salicylate. In (b, d): diagonal striped bars, OLO; horizontal striped bars, OLO + salicylate

Full size image

Hyperinsulinaemic–euglycaemic clamp

When evaluated by hyperinsulinaemic–euglycaemic clamp, insulin sensitivity tended to be decreased by oleate (n = 4), and was significantly decreased by OLO (n = 3) compared with saline (n = 4) (ESM Fig. 1).

Ex vivo studies in rat islets

Glucose-stimulated insulin secretion from islets isolated from rats i.v. infused with oleate or OLO was markedly impaired, but secretion was restored with salicylate (Fig. 3a, b). Oleate or OLO increased phosphorylated IκBα (Fig. 3c, d) and active nuclear NFκB (Fig. 3e, f), effects that were not apparent in the presence of salicylate.

Fig. 3
figure 3

Insulin secretory response to glucose (a, b), and quantification of phosphorylated IkBα (c, d) and active nuclear NFκB (e, f) in freshly isolated islets from fat-infused rats. Groups are as described for Fig. 1. In (a, b): saline, n = 16; oleate, n = 14; oleate + salicylate, n = 8; OLO, n = 12; OLO + salicylate, n = 6; salicylate only, n = 10. In (c–f): saline, n = 5–7; oleate, n = 4–6; oleate + salicylate, n = 4–5; OLO, n = 5–6; OLO + salicylate, n = 5–8; salicylate only, n = 4–6. Data are mean ± SEM. *p < 0.05, **p < 0.01 and ***p < 0.001 vs all (in a and b, comparison is to all groups at the same glucose concentration). White bars, saline; light grey bars, salicylate. In (a, c, e): black bars, oleate; dark grey bars, oleate + salicylate. In (b, d, f): diagonal striped bars, OLO; horizontal striped bars, OLO + salicylate

Full size image

Oleate increased the levels of islet mRNA for Il1b, Tnfa, Mcp1, Il1ra, Tgfb, Cd68 and Cox2; the effect of oleate was prevented by salicylate (Table 2). Interestingly, salicylate alone decreased these markers. The signal for mRNA of Inos (also known as Nos2; encoding inducible nitric oxide synthase [INOS]) mRNA was undetectable in all groups. Oleate (n = 6) increased islet ROS compared with saline (n = 5); however, ROS levels were not reduced by adding salicylate to oleate (n = 8), and salicylate alone had no effect (n = 4) (ESM Fig. 2).

Table 2 mRNA levels of inflammatory markers in freshly isolated islets of infused rats
Full size table

Hyperglycaemic clamp in mice

We used 48 h oleate infusion in mice, with or without BMS, which is a much more potent and specific IKKβ inhibitor than salicylate, at a dose previously found to inhibit IKKβ in vivo [14]. After the 48 h infusion and prior to the hyperglycaemic clamp the groups treated with oleate had higher plasma NEFA (oleate 1.205 ± 0.156 mmol/l; oleate + BMS 1.177 ± 0.136 mmol/l) than the groups infused with saline (0.761 ± 0.206 mmol/l) or BMS alone (0.689 ± 0.070 mmol/l). The glucose level was raised to 22 mmol/l in all groups (Fig. 4a). The GINF necessary to maintain the clamp was lower in oleate-infused animals but was similar to saline with oleate + BMS (Fig. 4b). Clamp insulin and C-peptide were not lower in oleate-treated mice than in control mice (Fig. 4c, d), in contrast to our oleate model in rats, but similar to our previous studies in mice [20] and our olive oil model in rats. Accordingly, the sensitivity index M/I [27] was lower in oleate-treated mice (Fig. 4e). Basal and clamp insulin and C-peptide levels were higher in the groups treated with BMS. BMS had no significant effect on M/I when added to oleate but on its own decreased M/I. DI was decreased with oleate infusion, whereas BMS completely prevented the oleate-induced decrease. BMS alone had no effect on DI (Fig. 4f).

Fig. 4
figure 4

Plasma glucose (a), GINF (b), plasma insulin (c), C-peptide (d), M/I (e) and DI (f) during hyperglycaemic clamps in mice treated for 48 h with: saline (n = 7); oleate, 0.4 μmol/min (n = 6); oleate + BMS, 0.12 mmol kg−1 day−1 (n = 5); or BMS only (n = 3). Data are mean ± SEM. **p < 0.01 vs all; † p < 0.05 vs saline; ‡ p < 0.05 vs oleate. White circles/bars, saline; black circles/bars, oleate; black triangles/dark grey bars, oleate + BMS; white triangles/light grey bars, BMS only

Full size image

Studies in beta cell-specific IKKβ-deficient mice

We also used Ikkb Δbeta cell and littermate floxed control mice to determine whether genetic silencing of IKKβ protects from fat-induced beta cell dysfunction ex vivo and in vivo. Ikkb deletion in islets of Ikkb Δbeta cell mice was confirmed by immunoblotting (Fig. 5a) and there was a suggestion of partial deletion in the hypothalamus (Fig. 5b). There was no significant difference in weight between ~13 week old control (29.7 ± 0.7 g, n = 14) and Ikkb Δbeta cell (28.5 ± 0.8 g, n = 13) mice. Oleate infusion for 48 h elevated plasma NEFA approximately threefold (Fig. 5c). Glucose-stimulated insulin secretion from islets isolated from oleate-infused controls was impaired and this impairment was prevented in Ikkb Δbeta cell mice (Fig. 5d).

Fig. 5
figure 5

IKKβ protein levels in islets (a) and hypothalamus (b) of Ikkb Δbeta cell mice and floxed controls (Ikkb F/F). Each lane represents islets pooled from two mice of the same genotype (a) or one hypothalamus taken from an individual mouse (b). Plasma NEFA (c) and insulin secretory response to glucose of freshly isolated islets of Ikkb F/F and Ikkb Δbeta cell mice (d) following 48 h oleate (0.4 μmol/min) or saline infusion. Control + saline, n = 8; control + oleate, n = 7; Ikkb Δbeta cell + oleate, n = 5; Ikkb Δbeta cell + saline, n = 9. Data are mean ± SEM. † p < 0.05 vs control + saline. White bars, control + saline; black bars, control + oleate; dark grey bars, Ikkb Δbeta cell + oleate; light grey bars, Ikkb Δbeta cell + saline

Full size image

Before hyperglycaemic clamping, NEFA were elevated approximately twofold in the oleate-infused groups (Fig. 6a). During hyperglycaemic clamps, the decreased M/I found in oleate-infused controls was not observed in Ikkb Δbeta cell mice (Fig. 6b), which may be explained by partial deletion of hypothalamic Ikkb driven by RIP2. Importantly, Ikkb Δbeta cell mice were protected from the decrease in DI induced by oleate in controls (Fig. 6c).

Fig. 6
figure 6

NEFA levels prior to hyperglycaemic clamp (a), M/I (b) and DI (c) during hyperglycaemic clamps in Ikkb F/F (control) and Ikkb Δbeta cell mice following 48 h oleate (0.4 μmol/min) or saline infusion. Control + saline, n = 8; control + oleate, n = 7; Ikkb Δbeta cell + oleate, n = 8; Ikkb Δbeta cell + saline, n = 5. Data are mean ± SEM. * p < 0.05 vs control + saline and Ikkb Δbeta cell + saline; †† p < 0.01 vs control + saline; ‡ p < 0.05 and ‡‡ p < 0.01 vs Ikkb Δbeta cell + oleate. White bars, control + saline; black bars, control + oleate; dark grey bars, Ikkb Δbeta cell + oleate; light grey bars, Ikkb Δbeta cell + saline

Full size image

In vitro studies in islets

To completely rule out that the effect of salicylate or BMS on beta cell dysfunction was mediated by systemic effects, we performed in vitro studies. Rat islets were cultured for 48 h in control or oleate-containing medium, with or without salicylate. Oleate decreased glucose-stimulated insulin secretion, and salicylate prevented this decrease (Fig. 7a).

Fig. 7
figure 7

Levels of insulin secretion (a), IκBα (b) and phosphorylated AMPKα level (c) in islets exposed for 48 h to control or oleate with/without salicylate. Control, 0.5% NEFA-free BSA in medium: (a) n = 7, (b) n = 7, (c) n = 8. Oleate, 0.4 mmol/l in 0.5% NEFA-free BSA: (a) n = 14, (b) n = 8, (c) n = 7. Oleate + salicylate, 0.25 mmol/l: (a) n = 10, (b) n = 7, (c) n = 8. Salicylate only, in BSA: (a) n = 7, (b) n = 4, (c) n = 6. (d) Insulin secretion in islets exposed to oleate for 48 h with/without BMS. Control, n = 7; oleate, n = 14; oleate + BMS, 3 μmol/l, n = 10; BMS only, in BSA, n = 7. Data are mean ± SEM. ** p < 0.01 vs all; † p < 0.05 vs control. White bars, control; black bars, oleate. In (a–c): dark grey bars, oleate + salicylate; light grey bars, salicylate. In (d): dark grey bars, oleate + BMS; light grey bars, BMS. CON, control; OLE, oleate; SLY, salicylate

Full size image

Oleate decreased total IκBα (a marker of IKKβ activity as phosphorylated IκBα is degraded) and the decrease was prevented with salicylate (Fig. 7b). Salicylate is known to activate AMPK by preventing its dephosphorylation [29] and AMPK activation can result in IKKβ inhibition [30], though salicylate has also been reported to directly inhibit IKKβ [11]. There was a tendency for oleate to decrease phosphorylation of AMPK, which appeared to be prevented with salicylate, but there was no significant effect (Fig. 7c). BMS, which inhibits IKKβ in AMPKα-null cells [31], prevented oleate-induced beta cell dysfunction similar to salicylate (Fig. 7d).

The IKKβ/NFκB pathway can mediate oleate-induced beta cell dysfunction by at least three mechanisms: (1) impairment of beta cell insulin signalling via serine phosphorylation of IRS [6]; (2) increase in COX-2-derived PGE2 [32]; and (3) production of NO through induction of INOS [33]. However, Inos mRNA was undetectable in our islets in the ex vivo studies. Oleate increased ser307-phosphorylated IRS-1 in cultured islets, an effect prevented by salicylate (Fig. 8a). Oleate also increased PGE2 release in media and this was prevented by salicylate (Fig. 8b). We also treated islets with the COX-2 inhibitor SC-236, which prevented the secretory defect induced by oleate (Fig. 8c), whereas the COX-1 inhibitor SC-560 did not (insulin secretion values at 22 mmol/l glucose relative to control: control = 1.00 ± 0.33 (n = 4); oleate = 0.59 ± 0.17 (n = 4); oleate + SC-560 = 0.14 ± 0.06 (n = 3); SC-560 = 0.51 (n = 2).

Fig. 8
figure 8

Serine phosphorylated IRS-1 (a) and PGE2 levels (b) in cultured islets exposed for 48 h to oleate or vehicle control with/without salicylate. Control, 0.5% NEFA-free BSA in medium: (a) n = 5, (b) n = 5. Oleate, 0.4 mmol/l in 0.5% NEFA-free BSA: (a) n = 5, (b) n = 5. Oleate + salicylate, 0.25 mmol/l: (a) n = 5, (b) n = 6. Salicylate only, in BSA: (a) n = 5, (b) n = 5. (c) Insulin secretion in control or oleate with/without the COX-2 inhibitor SC-236 (10 μmol/l). Control, n = 11; oleate, n = 10; oleate + SC-236, n = 9; SC-236 only, in BSA, n = 9. Data are mean ± SEM. *p < 0.05 and **p < 0.01 vs all; † p < 0.05 vs control. White bars, control; black bars, oleate. In (a, b): dark grey bar, oleate + salicylate; light grey bars, salicylate. In (c): dark grey bars, oleate + SC-236; light grey bars, SC-236. CON, control; OLE, oleate; SLY, salicylate

Full size image

Discussion

We examined the effects of prolonged NEFA exposure with or without IKKβ inhibitors on beta cell function in vivo, ex vivo and in vitro. We used our in vivo models of lipotoxicity in rats [2, 15] and mice [20]. These are models of beta cell dysfunction, as beta cell mass is not decreased by 48 h fat infusion [20, 24] and apoptosis is not increased [34]. During hyperglycaemic clamps in vivo, both insulin and C-peptide levels were lower in rats treated with oleate, indicating reduced insulin secretion; reduced glucose-stimulated insulin secretion was also found in isolated islets ex vivo. With olive oil, glucose-stimulated insulin secretion ex vivo was reduced but insulin and C-peptide levels (indices of absolute insulin secretion) during hyperglycaemic clamps were unaffected. The different absolute insulin secretion between oleate and olive oil is explained by the effect of olive oil to induce a greater degree of insulin resistance. This was demonstrated using hyperinsulinaemic–euglycaemic clamps, which are the gold-standard assessment for insulin sensitivity in vivo. In vivo, the beta cell compensates for insulin resistance by increasing secretion. In the absence of insulin resistance, absolute insulin secretion corresponds to DI. However, in the presence of insulin resistance, DI rather than absolute insulin secretion should be taken as a measure of beta cell function, which includes the ability of the beta cell to compensate for insulin resistance. DI was impaired by both oleate and olive oil, showing a decrease in beta cell function with both types of fat, consistent with the ex vivo results with islets. The different effects of oleate and olive oil on insulin sensitivity may reflect the amount of saturated fat in olive oil (16%) and/or the plasma triacylglycerol elevation induced by olive oil [35].

Our results showing that fat-induced beta cell dysfunction was prevented by salicylate in vivo and ex vivo in isolated islets suggest a role for inflammatory pathways involving IKKβ in lipid-induced beta cell dysfunction.

Inflammatory pathways are known to be activated in beta cell glucotoxicity [36] and may enhance lipotoxicity [37]. Our previous results show that oxidative stress plays a causal role in beta cell dysfunction induced by monounsaturated fat [2]. ROS are known activators of IKKβ which, in addition to phosphorylating IκBα and thereby activating NFκB, phosphorylates IRS, thus inhibiting insulin signalling. In our ex vivo study, salicylate prevented the increase in phosphorylated IκBα and nuclear active NFκB induced by oleate or olive oil, although, interestingly, salicylate alone had no effect at this dose, as previously seen in the liver [19]. An increase in ROS was induced by oleate, but salicylate did not prevent this effect. This suggests IKKβ is a downstream effector required for the previously demonstrated effect of oxidative stress to induce beta cell dysfunction [2]. IKKβ activation could also be unrelated to oxidative stress in the case of olive oil, which contains saturated fat. Saturated fatty acids can activate IKKβ via toll-like receptors (TLR) 2 and 4. Tlr4-null mice are protected from beta cell dysfunction induced by palmitate [38] and Tlr2-null mice are protected from beta cell dysfunction induced by high-fat diet [23]. Although saturated fat is believed to exert a more deleterious effect on beta cells than unsaturated fat, this is mostly based on in vitro data because, until recently, palmitate infusion has been a challenge [38].

Salicylate has previously been found to restore glucose-stimulated insulin secretion in an in vitro glucotoxicity model [7]. Although not all studies are concordant [39], the majority report that salicylate improves beta cell function in humans [40,41,42]. This effect was initially attributed to inhibition of COX-2 [32], the gene for which is controlled by NFκB, and the consequently decreased synthesis of PGE2, a prostaglandin which inhibits insulin secretion. In addition to inhibiting Cox2 transcription via NFκB, salicylate is a direct inhibitor of both COX-1 and COX-2 [43], and an activator of AMPK [29]. Salicylate did not likely protect against beta cell dysfunction through COX-1 inhibition as a COX-1 inhibitor did not prevent oleate-induced beta cell dysfunction. AMPK phosphorylation, however, did tend to decrease with oleate and this decrease was prevented by salicylate, which raises the question as to whether the protective effect of salicylate was mediated in part by AMPK activation. The effect of AMPK on insulin secretion is generally considered to be inhibitory [44]; however, AMPK may also deplete islet fat [45] and inhibit IKKβ [30]. To further implicate IKKβ in the effect of oleate to decrease beta cell function, we also used BMS, an inhibitor that, to our knowledge, has not been reported to activate AMPK; we obtained the same results for beta cell function as with salicylate both in vitro and in vivo. Importantly, the effect of IKKβ to mediate fat-induced beta cell dysfunction is also supported by our ex vivo and in vivo data using a genetic inhibition model, the Ikkb Δbeta cell mouse.

In mice, however, two important differences from the rat studies were noted. First, 48 h exposure to oleate induced marked insulin resistance in mice in accordance with our previously published data [20]. Besides species difference, the reason behind this finding may be the strain of mice, as C57BL/6 mice are very susceptible to fat-induced insulin resistance [46]. Oleate infusion, which induced insulin resistance in mice, did not result in lower absolute insulin secretion during the clamp, but decreased beta cell function (DI), similar to our results with olive oil in rats. Second, BMS did restore DI but did not affect oleate-induced insulin resistance, as it likely effected a decrease in insulin sensitivity itself, which resulted in increased plasma insulin and C-peptide during clamps. Decreased insulin sensitivity with an IKKβ inhibitor may be dose related and due to the inhibition of COX-2-derived prostaglandins that increase insulin action [47].

The mechanisms whereby IKKβ inhibition prevents fat-induced beta cell dysfunction deserve further study. However, both upregulation of insulin signalling and COX-2 inhibition are plausible mechanisms, as suggested by two lines of evidence. First, salicylate prevented serine phosphorylation of IRS, which is known to decrease insulin-induced tyrosine phosphorylation (i.e. insulin signalling). Beta cell insulin signalling is known to be important for beta cell function [48]. Second, salicylate prevented the oleate-induced PGE2 production and a COX-2 inhibitor mimicked the effect of salicylate. Previously, the effect of COX-2 on beta cell function has been studied mainly in the context of cytokine exposure, with contrasting results, presumably due to COX-1 vs COX-2 specificity of the inhibitors used [32, 49]. Also, the effects of exposure to COX-2 products yielded variable results among laboratories [49, 50] but a dose-dependent inhibitory effect [49, 50] suggests COX-2 may be implicated in decreasing beta cell function. As oleate increased Il1ra gene expression to a greater extent than that of Il1b, the increase in Cox2 mRNA expression is likely due to an oxidative stress-induced activation of IKKβ, independent of IL-1β. Nonetheless, salicylate did prevent upregulated gene expression of cytokines and chemokines, which may have contributed to beta cell dysfunction by further activating IKKβ and possibly other inflammatory pathways.

In summary, we demonstrated that prolonged exposure to fatty acids, which induces oxidative stress in islets, decreases beta cell function both in vitro and in vivo via activation of IKKβ. The novelty of our findings is the demonstration that IKKβ mediates beta cell dysfunction induced by NEFA selectively, and that the IKKβ/NFκB pathway is a therapeutic target to prevent NEFA-induced beta cell dysfunction in vivo.